DOCUMENTATION AND LEARNING

Start with the research task, then check the evidence.

These guides point to the workbench paths that exist today. Each one names the records it uses and the question it cannot answer for you.

New to the workspace?

Open the demonstration model and inspect one declaration before you run it.

Understand the demonstration

Have a source already?

Open the full workspace to import, inspect, or validate the file you hold.

Open the workspace

Find a task or term

Searches this page only; it does not inspect your model or send a request.

TASK GUIDES

Choose one bounded next step

Every link opens the matching task route in the current workbench.

Understand a model

Read the source, map the declared structure, and keep each explanation tied to an exact model element.

Evidence used: Open source, parsed model, and source-linked declarations
Limit: An explanation does not add outside biological facts or validate a model.
Owner: Model workbench
Update when: When parser, evidence, or assistant contracts change
Keep current: Open the source and follow one cited declaration.

Open this task

Reproduce a bounded run

Start with the demonstration model, preserve its input choices, and inspect the recorded result reference.

Evidence used: Exact source, selected inputs, solver settings, and verified run receipt
Limit: A completed run is numerical output, not a reproduced biological finding.
Owner: Run jobs and reproducibility export
Update when: When run planning, result receipts, or export manifests change
Keep current: Compare the receipt, source reference, and inputs before interpreting a result.

Open this task

Change one assumption

Locate the declared parameter or equation, prepare a precise change, then review the deterministic checks and diff.

Evidence used: Target declaration, proposed diff, capability checks, and approval record
Limit: A proposal does not write source or establish scientific suitability.
Owner: Source editing and proposal review
Update when: When the reviewed-change or approval flow changes
Keep current: Confirm the target, diff, and preconditions in the workspace.

Open this task

Compare models or runs

Keep sources, configuration, and compatibility limits beside the values you compare.

Evidence used: Selected source/run records and compatibility checks
Limit: A visual or numerical difference does not identify a biological cause.
Owner: Comparison workspace
Update when: When comparison or compatibility rules change
Keep current: Verify that both records remain comparable before drawing a conclusion.

Open this task

Review a composition

Check declared interfaces, units, and missing couplings before treating two components as a candidate system.

Evidence used: Component interfaces, mappings, and declared assumptions
Limit: Composition is review-only until a supported workspace can apply and execute it.
Owner: Composition contracts
Update when: When interface, mapping, or composition capability changes
Keep current: Read every unresolved coupling and unit diagnostic.

Open this task

Prepare a reproducibility record

Collect the source, run context, provenance, and export details needed for another person to inspect the work.

Evidence used: Verified source/run references, provenance, and export manifest
Limit: A complete record supports inspection; it is not peer review or scientific validation.
Owner: Export and publication records
Update when: When export manifests or provenance fields change
Keep current: Open the export preview and inspect its retained source and declared losses.

Open this task

EDITORIAL PLAN

Keep every guide attached to a product boundary

Each material has a named audience, task, evidence source, owner, review date, and stable product route. We do not publish customer stories or results without attributable user evidence.

Material	Audience and task	Evidence and owner	Reviewed	Product route
Start with the demonstration	Researchers new to the workspace Inspect a declared model element before a run	Built-in source, declared values, and visible provenance Model workbench	2026-07-20	Open route
Import a model safely	Researchers with a local SBML file Inspect import preflight and blockers	Original source bytes, parser diagnostics, and capability report Import recovery and format capabilities	2026-07-20	Open route
Run one bounded experiment	Researchers checking a declared scenario Preserve inputs and inspect a run receipt	Source, inputs, solver settings, and verified receipt Run jobs and reproducibility export	2026-07-20	Open route
Interpret evidence	Researchers reviewing a model or result Separate source, provenance, capability, and validation claims	Source-linked declarations, health report, and capability records Model workbench	2026-07-20	Open route
Export a reproducibility record	Researchers preparing a handoff Review retained source, settings, and declared losses	Verified source/run references and export manifest Export and publication records	2026-07-20	Open route

FORMAT SUPPORT

Read the operation boundary, not a format logo

The SBML row below comes from the same capability preflight used in the workbench. Other rows name their current source-bound contracts.

Format	Now	Execution	Export and loss
SBML Level 3 Version 2 Core	Inspect the parsed Core subset. Source-only constructs stay visible, and blockers are listed before a run.	Supported in this sample: Compartments, Species, Global parameters, Unit definitions, Rate rules.	The reproducibility bundle includes the exact open SBML source; Nendera does not rewrite it from the parsed model.
CellML 1.x and 2.0	Inspect retained source, components, interfaces, imports, and review drafts in the component workspace.	Not available from the CellML component review workspace.	Exact retained-source export is available; imports are not resolved and review drafts do not write CellML.
NeuroML 2 and LEMS	Source-bound structural inspection is available through the document contract.	Not available: no pyNeuroML, jNeuroML, or LEMS backend is selected or run.	Only unchanged source is retained by the document contract.
PBPK setup records	Read-only, source-bound setup and evidence planning is available.	Not available: no PBPK solver or dosing simulation is selected by the current contract.	PK-Sim import/export and a browser editor are not available.

REFERENCE

Records, limits, and terms

Reproducibility

Use the source, declared inputs, solver settings, and verified run record together. A ZIP bundle keeps those records inspectable; it does not independently rerun or validate a study.

Open the run path

Model provenance

Built-in samples and imported records carry different origin evidence. Check the source name, retrieval details, and checksum before relying on a claimed origin.

Inspect model evidence

Solver and analysis limits

The deterministic runner supports a bounded MathML and SBML subset. The health and capability views name constructs that remain source-only or block execution.

Open model health

Assistant safety

Codex can explain the open source and current run, then prepare reviewable proposals. Evidence chips resolve to the workspace or say when a record is unavailable.

Open the assistant

Exports

Export previews show the retained source, manifest, and declared transformations or losses before a bundle is created.

Prepare an export record

Glossary

Source: The exact model text currently opened in the workspace. A parsed view or export should identify which source it represents.
Provenance: Where a model record came from, such as a built-in sample or a checksum-verified import. It records origin; it does not judge quality.
Capability: A specific operation the current format and model can support, preserve as source-only, or block. Capability is narrower than a marketing promise.
Run receipt: A server-owned record that ties a completed run to its result reference and workspace state. It helps inspect execution lineage.
Proposal: A reviewable, non-applying candidate change. The separate workspace still owns deterministic checks and approval.
Scientific validation: Evidence that a model or result is fit for a scientific use. Parsing, unit checks, execution, and provenance each answer narrower questions.